Matthew Perry

asking on #gdal freenode right now. For the most part, I'd think that a libgdal compiled `--with-threads=no` is rare enough that testing under those environments could be accomplished with a...

For reasons outlined in #853 (TLDR, there are some edge cases wrt invertable coordinates which can mess up results), I think we should avoid using `reproject` and perform the resampling...

What about drivers other than GeoTIFF? - If the format supports internal overviews, there is likely a format-specific variable relevant to each driver. - Some formats, external overviews is the...

Noting that even if the `TIFF_USE_OVR` is set to `TRUE`, the original file is still altered by being opening in `r+` mode. ``` $ ls -l test.* -rw-r--r-- 1 mperry...

Interesting! I'd like to do some benchmarking in C to get a sense for the difference in performance and to understand which formats/creation options would benefit. Based on this mailing...

> This could be taken a step further and rasterio could instead (or aswell) provide an apply function... The user-defined python function would have to have a standard interface, accept...

Three follow up thoughts: 1) We might start with enforcing a key ordering for the rio info output. This would at least make the json output consistent and allow easier...

@dnomadb lets get a PR going for this. We've have to think a bit more about cell size/height and width but having it wrapped in the `merge_default_transforms` makes it easy...

This won't break anything in the existing rio warp usage so I'm calling it an enhancement/new feature that can wait until post 1.0

@grovduck I'm imagining that the questions of available memory and such would be handled by other parts of the application, with rasterio just providing a simple API to grab some...